You are on page 1of 3

EBM Learning: General medicine

BMJ EBM: first published as 10.1136/bmjebm-2018-111115 on 19 December 2018. Downloaded from http://ebm.bmj.com/ on 20 December 2018 by guest. Protected by copyright.
Inclusion and exclusion criteria and the problem of
describing homogeneity of study populations in
clinical trials
Franz Porzsolt,1,2 Felicitas Wiedemann,1,3
Susanne Isabel Becker,1,4 C J Rhoads5

10.1136/bmjebm-2018-111115 Introduction ►► Explicit and clear description of the study hy-


The description of results in clinical trials is an pothesis.
unsolved problem.1 The low quality of some clin- ►► Differentiation of primary and secondary end-
►► Additional material is
ical trials and the low output of useful data for points.
published online only. To
healthcare decisions is a problem for clinicians ►► Categorisation as randomised or non-ran-
view please visit the journal
online (http://​dx.​doi.o
​ rg/​ and patients alike. Statisticians doing system- domised trial.
10.1
​ 136/​bmjebm-2 ​ 018-​ atic reviews or meta-analysis try to solve this ►► Explicit and clear differentiation of the ex-

111115). problem by establishing lower p-value thresholds pected outcome described as efficacy or ef-
(for example, changing from 0.05 to 0.005). That fectiveness.
1
Institute of Clinical results in the acceptance of fewer, larger and more ►► Description of the study design as pragmatic
Economics (ICE) e.V., Ulm, or other than pragmatic trial.
carefully designed studies in order to be consid-
Germany
2 ered to have sufficient power.2 The problem is that ►► Description of the QoL instruments used.
Health Care Research at the
good useful information might get suppressed ►► Numbers of described inclusion and exclusion
Department of General and
Visceral Surgery, University using stricter p-value thresholds. This is an espe- criteria.5
Hospital Ulm, Ulm, Germany cially difficult problem in quality of life (QoL) In August 2016, Wiedemann conducted
3
Klinik für Allgemein- und studies which, by their very nature, are smaller a PubMed search with the following search
Viszeralchirurgie, Diakonie- and less powerful. parameters:
Klinikum Stuttgart, Stuttgart, QoL researchers believe that the answer to poor ►► The terms ‘trial’ and ‘quality of life’ in their
Germany quality is to review more studies that describe titles.
4
Klinik für Seelische meaningful positive outcomes that contribute to ►► Free full text available.
Gesundheit, Karl Hansen The number of articles with trial and quality of
the improvement of healthcare.3 What is often
Klinik, Bad Lippspringe,
missing is a general discussion about the selection life in the title for the past 5 years is depicted in
Germany
5 criteria. figure 1. Based on this information, it was reason-
College of Business,
Kutztown University, Choosing selection criteria is not easy. On one able to choose to review the first 100 free acces-
Kutztown, Pennsylvania, USA hand, wide inclusion criteria should be chosen in sible articles that were listed as of August 2016.
order to demonstrate safety and effectiveness for Papers were downloaded as PDFs and analysed.
a large group of patients. On the other hand, there Subsequently, some of the articles were updated
Correspondence to: are many ethical and scientific reasons to define and/or no longer freely available. However, in all
Dr Franz Porzsolt, Health exclusion criteria.4 The resulting subjective selec- cases, the inclusion and exclusion criteria of the
Care Research at the Dept tion of inclusion and exclusion criteria should be earlier version and the later version were identical.
General and Visceral Surgery, The analysis was not impacted by the subsequent
a balance of these two considerations.
University Hospital Ulm, Ulm
In this report, we analyse the information updates. The complete list of 100 articles can be
89071, Germany; f​ .​porzsolt@​
provided by the description of inclusion and found in online supplementary appendix A.
gmx.d​ e
exclusion criteria. We demonstrate that valuable In this report, we focus only on the numbers
information on the investigated study popula- and types of described inclusion and exclusion
tion can be provided by a complete and precise criteria, and leave the other six criteria for future
description of the inclusion and exclusion criteria reports.
that is not present in all publications. Furthermore,
we propose that researchers can derive important Results
© Author(s) (or their information on the homogeneity/heterogeneity Fifteen of the 100 documents were only study
employer(s)) 2018. No of different study populations from completely protocols and 85 were reports of completed
commercial re-use. See reported inclusion and exclusion criteria. studies. Fourteen were published in 2017, 64 in
rights and permissions. 2016 and 22 in 2015.
Published by BMJ. Method Both inclusion and exclusion criteria were
To cite: Porzsolt F, This study builds on on the data of research done reported in 69 of 100 publications. Only inclusion
Wiedemann F, Becker SI, by Wiedemann5 which analysed seven charac- but no exclusion criteria were reported in six. There
et al. BMJ Evidence-Based teristics of 100 publications on QoL assessments. was no paper that reported only exclusion but no
Medicine Epub ahead of Wiedemann identified factors which contributed inclusion criteria, but there were 25 publications
print: [please include Day to the aforementioned problem of poor results that reported neither inclusion nor exclusion
Month Year]. doi:10.1136/ reporting for clinical trials. The seven analysed criteria. Some of these 25 publications described
bmjebm-2018-111115 factors were: the selection criteria ‘briefly’ or mentioned only

BMJ Evidence-Based Medicine Month 2018 | volume 0 | number 0 | 1


EBM Learning: General medicine

BMJ EBM: first published as 10.1136/bmjebm-2018-111115 on 19 December 2018. Downloaded from http://ebm.bmj.com/ on 20 December 2018 by guest. Protected by copyright.
Figure 1: Number of PubMed articles with Trial and Quality of Life in their
title by year.

Figure 2: Number of Studies with Number of Selection Criteria


‘key inclusion criteria’. All of these 25 documents claimed that
the precise inclusion and exclusion criteria had been described
in previous publications. Unfortunately, the references of some selection criteria is essential for appropriate interpretation of the
of these publications were not provided and in other cases it was study results.
necessary to check several referring references until the document
describing the relevant information could finally be identified. Lack of precise definitions
These results are shown in table 1. Both the inclusion and exclusion criteria can each represent
Inclusion criteria were reported in detail in 75 of 100 publica- one of the two endpoints of a dichotomous selection criterion.
tions. The median number of explicitly reported inclusion criteria Inclusion criteria describe the conditions a patient has to meet
in these 75 documents was 4 with an IQR of 3 to 5. Twenty-four to be included in a study and the exclusion criteria describe the
of the remaining 25 publications mentioned the exact reference conditions a patient must not meet to be included in a study. The
that describes the inclusion criteria. One document mentioned problem is that the same criteria can be named inclusion (patients
the previous description of the inclusion criteria but did not with a particular disease are included) or exclusion criteria
provide a traceable reference. Exclusion criteria were described (patients without this particular disease are excluded).6 7 Criteria
in 69 of 100 publications, the median being 4 with an IQR of 2 should be specified as inclusion or exclusion.8 9
to 6. Of the remaining papers, 25 provided a reference for the
exclusion criteria, and 6 did not mention any exclusion criteria Waste of information
at all. Figure 2 shows the number of studies of the 75 studies that The quantitative and qualitative information of the inclusion
reported one or more selection criteria. The number of reported and exclusion criteria could be used as a description of the target
selection criteria varied, with two studies reporting more than 20 population. The total number of inclusion and exclusion criteria
selection criteria. Six studies reported only inclusion criteria. The can be used as a simple measure for description of the homoge-
median number of selection criteria per study was 8 with an IQR neity of a target population. The more criteria used to define the
of 5 to 10. study population, the more homogeneous the resulting popula-
tion. Our data show that the populations included in our investi-
Discussion gated studies were quite heterogeneous.
Our results confirm that precise inclusion and exclusion criteria The total number of selection criteria (ie, the sum of the inclu-
were described in 69% of the investigated publications. But there sion and exclusion criteria) as indicator of the homogeneity of the
are also three negative items: (1) incomplete reporting, (2) lack of investigated population is only a very rough and lossy proxy for
a precise definition and (3) useful information wasted. the underlying selection. Each of these criteria may have a strong
or weak effect on the specific outcome investigated in a partic-
Incomplete reporting ular study. Each criteria may be dependent or independent from
If the reader cannot find the information about the inclusion each other. So even if the same numbers of selection criteria are
criteria, then the information provided is incomplete. In 25% used, the degree of homogeneity will be influenced by the types
of the investigated publications, neither inclusion nor exclusion of selection criteria. Nonetheless, having this information is useful
criteria and in 6% no exclusion criteria were directly reported and can help a researcher determine whether or not two different
with references that could be found. Complete reporting of all studies are comparable.

Conclusion
Table 1: Number of Inclusion a d Exclusion Criteria Reported We make three concluding recommendations from our research:
Inclusion criteria 1. Report selection criteria completely.
Reported Not reported Total 2. Define inclusion and exclusion criteria separately.
3. Clearly differentiate between inclusion and exclusion criteria.
Exclusion criteria
Our recommendations would make it easier to replicate clin-
 Reported 69 0 69
ical studies in the real world. Exclusion criteria are only appli-
 Not reported 6 25 31
cable to ideal study conditions (ISCs) because some patients would
Total 75 25 100
influence the assessed study outcomes. Inclusion criteria can be

2 BMJ Evidence-Based Medicine Month 2018 | volume 0 | number 0 |


EBM Learning: General medicine

BMJ EBM: first published as 10.1136/bmjebm-2018-111115 on 19 December 2018. Downloaded from http://ebm.bmj.com/ on 20 December 2018 by guest. Protected by copyright.
included when the study is replicated under real-world conditions error and fraud in research underpinning patient care357 doi. Evid Based
(RWCs).10 11 Med 2017;22.
In addition, there are ‘research methods’ such as randomisation 2 Ioannidis JPA. The Proposal to Lower P Value Thresholds to.005.
or blinding that are needed for completion of high-quality exper- JAMA 2018;319:1429–30.
3. Bullinger M. [The concept of quality of life in medicine: its history and
imental trials but are not used under RWC studies.
current relevance]. Z Evid Fortbild Qual Gesundhwes 2014;108:97–103.
A clear distinction of inclusion and exclusion criteria is helpful
4. U.S. Food and Drug Administration. Evaluating Inclusion And Exclusion
for the exact differentiation of efficacy and effectiveness. Efficacy
Criteria. Clinical Trials: Workshop Report. Washington, DC, 2018. https://
describes effects under ISC while effectiveness describes effects www.​fda.​gov/​downloads/​Regu​lato​ryIn​form​ation/​LawsEnforcedbyFDA/​Sign​
under RWC. Efficacy and effectiveness cannot be differentiated ific​antA​mend​ment​stot​heFDCAct/​FDARA/​UCM613054.​pdf.
without clear description of inclusion and exclusion criteria. 5. Wiedemann F. Potential mistakes in clinical trials that use health-related
quality of life as outcome: Thesis at the Medical Faculty University of Ulm,
In summary 2018.
In summary, simple modifications to the clarity with which selec- 6. Williams EM, Lorig K, Glover S, et al. Intervention to Improve Quality
tion criteria are described in published clinical trials would go a of life for African-AmericaN lupus patients (IQAN): study protocol for a
long way towards the goal of improving the quality and useful- randomized controlled trial of a unique a la carte intervention approach
ness of the study conclusions in QoL studies. to self-management of lupus in African Americans. BMC Health Serv Res
2016;16:339.
7. Omranifard V, Rajabi F, Mohammadian-Sichani M, et al. The effect of
Contributors All authors contributed substantially to the paper.
add-on memantine on global function and quality of life in schizophrenia:
FW and FP were primarily responsible for the acquisition of the
a randomized, double-blind, controlled, clinical trial. Adv Biomed Res
data, but all authors contributed substantially into the analysis 2015;4:211–5.
and review of the data, the development of the report and editing 8. Pasquau J, Hidalgo-Tenorio C, Montes ML, et al. High quality of life,
of the paper for submission. treatment tolerability, safety and efficacy in HIV patients switching from
Competing interests None declared. triple therapy to lopinavir/ritonavir monotherapy: a randomized clinical
trial. PLoS One 2018;13:e0195068.
Ethics approval This study includes only published information 9. Moretti HD, Colucci VJ, Berry BD. Vitamin D3 repletion versus placebo as
that cannot be related to individual subjects. adjunctive treatment of heart failure patient quality of life and hormonal
Provenance and peer review Not commissioned; externally peer indices: a randomized, double-blind, placebo-controlled trial. BMC
reviewed. Cardiovasc Disord 2017;17:274.
10. Porzsolt F, Rocha NG, Toledo-Arruda AC, et al. Efficacy and effectiveness
Data sharing statement The results on the remaining six criteria trials have different goals, use different tools, and generate different
are available (in German language) through FW and FP. We are messages. Pragmat Obs Res 2015;6:47–54.
planning additional publications on these papers. 11. Porzsolt F. The assessments of three different dimensions “efficacy”,
“effectiveness”, and “value” require three different tools: the randomized
References controlled trial (RCT), the Pragmatic Controlled Trial (PCT), and the
1. Heneghan C, Mahtani KR, Goldacre B, et al. Evidence based medicine Complete Economic or Cost-Effectiveness Analysis (CEA). Surgery and
manifesto for better healthcare: a response to systematic bias, wastage, Rehabilitation;2.

BMJ Evidence-Based Medicine Month 2018 | volume 0 | number 0 | 3

You might also like